A Device Anomaly Detecting Method Based On Event Correlations Among Multi-source Event Streams

With the advent of the information age,technology has developed rapidly and sensors have been widely used in industrial environments.By processing and analyzing various running state data of the device collected by the sensors,anomaly detection of the device can be carry out.The industrial environment requires high accuracy and real-time performance of anomaly detection methods.And industries hope to minimize or avoid the losses caused by the anomalies.But the traditional anomaly detection of device are mostly applied to static data sets,and it is difficult to get ideal results on the sensor stream data generated in the actual industrial environment.Based on the analysis of multi-source sensor data,this paper explores the co-occurrence correlation between abnormal events,proposes an online detection method of equipment abnormality,and the guarantee methods for the accuracy and immediacy of anomaly detection is studied to meet the needs of industrial equipment abnormal detection.The main research is summarized as follows:1.Aiming at the problem of discovering correlation in multi-source anomaly event stream,a method of instant discovery of event correlations in multi-source event streams based on frequent co-occurrence pattern mining is proposed.The method is based on sliding window model.By counting event in the multi-source event streams,a frequent co-occurrence pattern mining method with time series features is proposed to accurately and efficiently discover event correlations.2.Aiming at the problem of anomaly detection of industrial equipment,an online anomaly detection method based on event correlations is proposed.The method can access the sensor data in real time,based on the mining event correlations can predict the subsequent abnormal events with the greatest probability,and detect anomalies in real time.3.Based on the real sensor data collected by a thermal power plant,a lot of experiments are carried out to verify the accuracy and performance of the proposed method based on multi-source event flow event correlations.4.Based on the instant discovery method of event correlations and anomaly detection method proposed in this paper,a prototype system named device health condition diagnosis system is designed and implemented.This system can monitor the operation of power plant equipment in real time,support online abnormal detection of equipment,and diagnose th e running state of equipment.